Psychodynamic and Cognitive-Behavioral Therapies Are More Different Than You Think: Conceptualizations of Mental Problems and Consequences for Studying Mechanisms of Change

Essentialist model

The essentialist model assumes that kinds have essences—that is, sets of features necessary and sufficient for something to count as a member of that kind. Moreover, from the essential sets of features, many identifying characteristics of that kind are supposed to arise. A typical example of an essentialist kind is an element from the periodic table. The essence of an element is its atomic number, the number of protons in its nucleus. The surface properties (melting point, malleability, color, and resistance to oxidation) of, for instance, gold follow lawfully from its atomic number. Furthermore, gold can be identified by checking if that essence is present. Every atom of gold has 79 protons. No atom that is not gold has 79 protons.

In the widely used disease model of mental disorders, mental disorders are considered to be essentialist kinds. Analogous to the structure of somatic diseases (e.g., a virus causing the symptoms of flu), mental problems cluster together in disorders because they are caused by the same underlying process. This process may be of a genetic, biological, psychological, or developmental nature. It constitutes the essence of the disorder—that is, the necessary and sufficient condition for the disorder to exist. Consequently, it is possessed by all individuals with the disorder and no individuals without the disorder. All the typical features of a disorder, for instance the clinical symptoms and features of illness course of depression, are caused by this essential underlying process. Similar to the role of protons in the case of gold, the underlying cause makes a phenomenon what it is and distinguishes it from other phenomena. Moreover, it is assumed that treatment of such a common cause will relieve all the symptoms of the disorder. Therefore, the focus of treatment is preferably on the underlying cause and not on the symptoms.

MPC model

According to the alternative MPC model (Kendler et al., 2011), the co-occurrence of symptoms in disorders is maintained by the existence of causal mechanisms. These mechanisms ensure that symptoms are instantiated together and co-occur from individual to individual. Disorders are sets of symptoms connected through a system of causal relations. For instance, in depression, stressful life circumstances may lead to lack of sleep, which, in turn, leads to fatigue, which elicits concentration problems, which, in turn, feed back to and reinforce the sleeping difficulties, and so forth. Thus, symptoms may cause each other and their causal relations explain their co-occurrence; there is no need for postulating an underlying causal disease.

Measurement models

Testing explanations of mental disorders requires that the relevant clinical phenomena are somehow observed, described, and measured. The essentialist and MPC models presuppose different measurement models of the relationship between abstract psychological concepts called constructs (e.g., depression, neuroticism, intelligence) and observables. Observables include verbal and other behaviors; bodily expressions; mental activities such as attending to, thinking of, or imagining something; and experiences such as feelings (of anxiety, sadness, anger, joy) and visual, auditory, and bodily sensations. These phenomena are observable in the sense that they have genuine duration—that is, one can follow with attention when they start, go on, and finish (Wittgenstein, 1967, p. 45). Thus, these phenomena can be rather unambiguously identified, either by introspection ² or by behavioral observation. The phenomena referred to by constructs, on the other hand, manifest themselves in certain observable experiences and behaviors, without being these experiences or behaviors themselves.

When constructs are supposed to refer to real entities causing the observables, as is the case in essentialist explanations, a reflective measurement model is adopted (Borsboom, Mellenbergh, & van Heerden, 2003; Edwards & Bagozzi, 2000). In this model, constructs are called latent. The latent construct (e.g., depression) represents the common cause of the observable variables (e.g., feeling down, lack of sleep). The observable variables are regarded as indicators reflecting the essential disease process (depression itself) and are therefore aggregated in a total score that is supposed to show a person’s position on the latent construct.

MCP models are consistent with a relatively new psychometric approach, called network analysis, in which constructs are seen as causal networks of directly related observables (Hofmann, 2014; Hofmann, Curtiss, & McNally, 2016; Schmittman et al., 2013). The variables that are typically taken to be observable indicators of latent constructs are taken to be autonomous causal entities in a network. Thus, studying disorders means investigating their network structure—that is, the pattern of causal relationships between the observable symptoms. Disorders are supposed to exist as systems rather than as entities. Therefore, the aggregation of symptoms is of less interest from a network perspective. The focus of assessment is rather on the strength and number of connections between symptoms and the centrality of individual symptoms. For example, Bringmann, Lemmens, Huibers, Borsboom, and Tuerlinckx (2015) administered the Beck Depression Inventory–II (BDI-II; Beck, Steer, & Brown, 1996) weekly to depressed patients receiving either CT or interpersonal therapy (IPT) and analyzed the temporal dynamics of the symptoms. The analysis showed that all BDI-II symptoms were directly or indirectly connected through patterns of week-to-week influence. In addition, these influences were often mutually reinforcing. Loss of pleasure turned out to be the most central symptom in the network, being located most frequently on the paths between other symptoms as well as having the strongest influence on other symptoms the following week. Suicidal thought and thoughts of worthlessness had the strongest influence next to loss of pleasure.

The SET model of SAD

As defined in DSM-5 (American Psychiatric Association, 2013), SAD entails a marked fear of situations in which the individual is exposed to possible scrutiny by others. There, the individual fears that he or she will act in a way or show anxiety symptoms that will be negatively evaluated. According to SET (Luborsky, 1984), SAD develops and persists because it is part of a “transference potential” that arose in childhood and is reproduced repeatedly (Leichsenring et al., 2007, p. 60). The content of this potential is called the Core Conflictual Relationship Theme (CCRT; Luborsky, 1984). A CCRT consists of three components: a wish (W; sometimes called affect), an expected response from others (RO), and a response from the self (RS). The core conflict between W and RO determines and thus explains RS (Leichsenring et al., 2007, p. 59). The RS component represents the patient’s symptoms and defenses. Thus, the features of SAD are conceptualized as part of the RS component of the patient’s CCRT (see Fig. 1).

OPEN IN VIEWER

Fig. 1.

The SET model of SAD. An internal conflict constellation determines all the symptoms of SAD, symbolized by arrows from the conflict to the symptoms. The model accords with an essentialist view of mental disorders in that the conflict constitutes the causal essence from which the typical features of SAD arise.

Let us imagine a patient, Peter, who seeks treatment for SAD from a therapist using SET (Luborsky, 1984). A formulation of Peter’s CCRT is identified through a relationship episode (REP) interview, where the patient is asked to “tell some events involving you and another person.” In the REP interview, Peter tells about a recent episode at his work place. He had just started in this job and joined a coffee break with the boss and two colleges. The boss’s presence was particularly stressful. He felt that he shook visibly and that the others stared at him, noticed his shakiness, and probably thought he was a “nervous wreck.” After this incident, Peter went on a sick leave and sought therapy.

We could imagine that he also tells the following old one: “At my father’s birthday party, when I told him that I had successfully passed my exam, he only said ‘Your brother was faster, boy!’ I felt terribly ashamed” (an example from Leichsenring et al., 2007, p. 60).

On the basis of these and other REPs, the therapist could derive the following CCRT: “I wish to be affirmed by others (W). However, the others will humiliate me (RO). I feel ashamed and get afraid of being together with others, so I have decided to avoid exposing myself (RS)” (Leichsenring et al., 2007, p. 59). This is an explanation of behavior and emotional responses in terms of wishes, expectations, and a conflict between these: Peter avoids social situations because he wishes to be affirmed but expects to be humiliated (implicit wish to avoid). The expected humiliation by others would severely frustrate his strong wish to be affirmed by others as well as his wish not to be humiliated; therefore, it elicits anticipatory anxiety, anxiety within social situations, feeling of control loss, and feeling of being blocked (Leichsenring et al., 2007, p. 59). If humiliation is actually experienced, also shame is elicited. Thus, in a CCRT, behaviors and emotional responses involved in SAD are explained in terms of conflicting wishes and expectations.

Moreover, due to a lack of a solution and to avoid the anxiety associated with them, wishes for affirmation, wishes for a real relationship, sexual wishes, and anger are often repressed (Leichsenring & Salzer, 2014, p. 236). In some cases, anticipatory anxiety and avoidance may serve as a defense against these wishes (Leichsenring et al., 2007, p. 68). Thus, the presence and persistence of the SAD symptoms are partly explained by their defensive purpose. In this way, a complex “potential” is developed, consisting of conflicting wishes and expectations and a limited ability to verbalize and reflect over these. In certain situations, this potential is activated and expressed in symptoms and defenses (activities and experiences). Thus, behaviors and other responses are explained in terms of wishes and expectations, including a wish not to be aware of these wishes and expectations.

It is emphasized that SAD is a heterogeneous disorder. Subpopulations of SAD may be differentiated according to, for instance, the specific conflict constellation. Some may fear to appear incompetent due to their unrealistic demands of themselves. Others may fear to be perceived as being similar to a close person of whom they are ashamed (Leichsenring et al., 2007, p. 64). However, for individuals within each subgroup, the CCRT is supposed to be the same.

SET adopts an essentialist model of psychiatric kinds

The SET model accords with an essentialist view of mental disorders. An internal conflict constellation (a CCRT) consisting of wishes (affects) and expectations is said to determine and thus explain the defenses and the cluster of symptoms of SAD (RS component) (Leichsenring et al., 2007, p. 59). Repression of wishes, anxiety, avoidance, shame, feeling of control loss, and feeling of being blocked are all seen as parallel effects of the internal conflict. Thus, the internal conflict constitutes the essence from which the typical features of SAD arise. Although there are different subgroups of SAD individuals, the same essence in terms of a definite conflict constellation of wishes and expectations is supposed to determine the symptoms within each subgroup.

As essentialist explanations are associated with a reflective measurement model, it follows that the SET explanation conforms to a reflective model. A variety of concrete and observable defenses and symptoms are treated as indicators of the underlying causal entity—the core conflict. The concepts of the underlying causes are called latent constructs.

Moreover, in reflective models, the observable indicators are seen as exchangeable (Edwards & Bagozzi, 2000). In the CCRT example, the observable symptoms—anticipatory anxiety, anxiety within and avoidance of social situations, feeling of control loss, and feeling of being blocked—have all the same role as effects of the core conflict. Their limited role in treatment is to indicate or point to the causal conflict, and one symptom is as reliable as any other. Therefore, it does not make a qualitative difference which indicator one focuses on in treatment. Finally, the SE model accords with a reflective model in assuming local independence of the symptoms. There is no mention of causal relations between the symptoms. They are viewed as effects of the conflict, and no causal connections between them are proposed.

In addition to examining the propositions of the SET explanatory model, we should also investigate whether the therapeutic prescriptions of the model accord with the essentialist model. Leichsenring and Salzer (2014, p. 231) provide an overview of the targets of intervention in SET for anxiety disorders. One target is the patient’s lack of awareness of the CCRT associated with the symptoms. The goal is to attain new insight—that is, an understanding of the patient’s CCRT (W, RO, RS). This is achieved through analysis of reported REPs and interpretation of the fearful transference to the therapist (Leichsenring et al., 2007, p. 71). Also, exposure to anxiety-provoking social situations is encouraged to explore the underlying core conflict. In these cases, the main questions are about meaning in terms of wishes and expectations: “What does this activity/experience (e.g., behavior, focus of attention, thinking, imagery, image, feeling, sensation) mean?” For instance, the following is an example of exploring expectations: “You are afraid to tell your boss that you are angry with him. How would he react if you do so?” (Leichsenring et al., 2007, p. 73). Thus, the observable symptoms are taken as indicators of the features (wishes and expectations) of the underlying conflict or essence and used as means to identify these features. Questions about how the symptomatic activities and experiences may influence each other are not addressed.

Another target is an aspect of the RS component of the CCRT—the scaring internal dialogue. Here a so-called supportive intervention—fostering an encouraging inner dialogue—is used. This dialogue is purported to be helpful when the patient confronts an anxiety-provoking situation. The patient is asked, for example: “We have learnt that your anxiety formula makes you anxious. However, is there anything that you can say to yourself that would encourage you?” (Leichsenring & Salzer, 2014, p. 237).

We can conclude that, in line with an essential model of mental disorder, an underlying conflict is viewed as a common cause of the SAD symptoms. Therefore, the overall focus of interest in SET therapy is latent entities, namely the underlying and partly warded-off wishes (affects) and expectations to which the observable symptomatic and other activities (e.g., defenses) and experiences are supposed to point. It is assumed that treatment of such a common cause will relieve all the SAD symptoms. The local independence assumption of reflective models is consistently adhered to. Even when supportive interventions directed at observable symptoms (e.g., a scaring internal dialogue) are used, they are not based on an analysis of potential causal relations between observables. Rather, SAD patients are helped to simply do the opposite of the problematic activity (e.g., foster an encouraging internal dialogue).

The CT model of SAD

The cognitive model (Clark & Wells, 1995) explains the persistence of social anxiety in SAD by showing how the person’s processing activities, experiences, and behavior within social situations interact to hinder a realistic perception of one’s own appearance and other’s attitudes. The model is presented in Figure 2. When a person with SAD enters social situations, a number of anxiety reactions are elicited. First, thoughts of making a fool of oneself and being negatively evaluated by others are elicited, and these thoughts evoke somatic and behavioral symptoms of anxiety. Second, these symptoms become further sources of perceived danger and anxiety (e.g., sensation of blushing is interpreted as evidence that one is making a fool of oneself). Third, negative images occur in which SAD persons see themselves as they fear they might appear to others. Fourth, SAD persons become preoccupied with their somatic responses and negative thoughts and images, and this preoccupation strengthens the certainty of these thoughts and interferes with the person’s ability to process social information. Thus, they are less likely to discover the actual reactions of other people, which usually are less negative than they think. Fifth, to avert the feared outcomes from occurring, the person engages in safety behaviors (e.g., grasping a glass tightly to prevent shaking). However, performing safety behavior prevents the patients from learning that most of their feared outcomes would not have happened in any case. Sixth, most safety behaviors increase self-monitoring and self-focus and hence are likely to increase the certainty of negative thoughts and thus social anxiety. Seventh, performing safety behaviors (e.g., rehearsing what to say next when a conversation partner is talking) takes up attention. This may make the SAD person appear less warm and outgoing and thus elicit less friendly behavior from others. In this way, the patients’ fear may partly be confirmed. Finally, safety behavior and self-focus may directly increase the likelihood of feared predictions occurring (e.g., monitoring how one is coming across may make one come across less well and gripping glasses tightly may increase the chance that one’s hands will tremble; McManus et al., 2009).

OPEN IN VIEWER

Fig. 2.

The cognitive model of in-situation social anxiety in social anxiety disorder. The reactions of an imagined patient, Peter, are filled in. The model accounts for the persistence of social anxiety by depicting the causal interrelationships between observable activities and experiences. Thus, it accords with an MPC view of mental disorders. Adapted from Figure 4.1 on page 72 in Clark and Wells (1995). Adapted with permission.

Let us then imagine that Peter seeks treatment from a therapist using the cognitive model of SAD. Here the case formulation that guides treatment is an individual version of the cognitive model. This formulation is derived from an analysis together with the patient of a recent anxiety-provoking social episode. The episode identified for Peter is the one at the coffee machine. By the therapist asking bridging questions guided by the model (e.g., “When you thought you would make a fool of yourself, what did you feel?”), the individual model for Peter is derived (filled in Fig. 2). Peter’s appraisal of social danger, that he would shake visibly and be judged as weird, led to anxiety feelings and sensations, most important to a strong sensation of hands shaking. As he noticed himself becoming more anxious, he focused attention on the sensation of shaking and on a spontaneous image of himself with hands trembling and spilling coffee. When focusing on the sensations and image, he became more certain that he started to shake visibly. To limit the shaking, he gripped tightly around the cup, tensed muscles in arms when lifting the cup, and monitored closely the lifting movement. In addition, to prevent the others from noticing, he turned his back to them when drinking. Still, he felt that he shook visibly and that the others stared at him, noticed his shakiness, and probably thought he was weird.

Thus, the cognitive model depicts a series of causal feedback relationships (“vicious circles”) between processing activities, experiences, and behaviors. In general, cognitive behavioral models may be viewed as causal schemas for mental disorders, representing observable variables presumed to be relevant for the mental disorder in question and the causal relationships among these variables. These variables refer to phenomena with genuine duration, meaning that they can be reliably identified, either by introspection or by behavioral observation. Moreover, they arise, last, and abate over limited periods of time, often within seconds. Thus, they constitute the microlevel of mental and behavioral life.

The CT model is consistent with an MPC model

Instead of focusing the relationship between a presumed essence (a core conflict) and observable features (symptoms), the CT model accounts for social anxiety by analyzing the causal interrelationships between observables (activities and experiences). This accords with the MPC view of mental problems and disorders. The causal relations between the observed features of SAD are the reason why SAD appears and is perceived as an entity. Because of the causal relations, every feature of SAD makes the presence of (some of) the other features more likely.

Thus, the cognitive model can be seen to depict the network structure and dynamics of the construct SAD. Single symptoms (e.g., in-situation social anxiety, anticipatory anxiety) are analyzed into their constituents—that is, into observable inner and outer activities and experiences such as spontaneous imagery and performance of safety behaviors, and the functional relationships between these. Thus, a symptom is explained on the microlevel of momentary mental events and behaviors occurring in the flow of daily life. For instance, it shows that self-focus is a central causal agent in the structure. It also demonstrates how the individual inadvertently averts receiving corrective experiences of another person’s attitudes toward him/her. Most importantly, it shows the dynamics of how a state of social anxiety in social situations may shift to a different state, thus forming a basis for intervention.

The cognitive model also explains the network structure of other symptoms of SAD, such as anticipatory anxiety and postevent rumination. For instance, the causal structure and dynamics of anticipatory anxiety are depicted in the following way: Prior to a social event, persons with SAD typically review in detail what they think may happen. As they start to think about the situation, they become anxious and their thoughts tend to be dominated by recollections of past failures and by negative images of themselves in the situation. Moreover, they take their felt anxiety as evidence that their negative predictions will come true.

From the network analysis of the single symptoms, also the causal relationships between symptoms can easily be derived. For instance, anticipatory anxiety has left the person already in a self-focused processing mode when entering the anticipated social situation, thus already with negative perceptual biases, which makes him/her more likely to perceive signs of failure (e.g., own shivering). Thus, more anticipatory anxiety causes more social anxiety within the anticipated social situation.

To what extent are the therapeutic prescriptions of the CT model consistent with the MPC perspective? Once the patient and therapist have agreed upon a working version of the cognitive model, a number of exercises and procedures are used to explore the functional relationships between observables. For instance, focus of attention and safety behaviors are manipulated (Clark & Wells, 1995). Patients are asked to role-play a feared interaction under two conditions. In one condition, they are asked to focus attention on monitoring themselves and to use all of their normal safety behaviors. In the other condition, they are asked to drop their safety behaviors and focus their attention on the other person(s) and what is said. The purpose is to explore whether self-focus and use of safety behaviors lead to increased anxiety and experience of failure rather than their expected reduction. The contrast to SET is particularly salient in the use of social exposure. While SET patients seek out feared social situations to explore their reactions and relate them to their CCRT (Leichsenring et al., 2007, p. 73), the CT patients are encouraged to change their behavior within social situations and notice the effects of this on the other observables of their personal model. More specifically, the patients change activities that are supposed to maintain social anxiety based on the theoretical model—that is, the use of safety behaviors and self-focus. Thus, they drop safety behaviors such as trying to look at ease and focus on other people and on what they are doing together with them. The focus of interest is not what these activities and experiences mean—as it is in SET—but how they influence and maintain each other. The main questions in CT are therefore about temporal sequences and causal relationships between activities/experiences: “What activities/experiences (e.g., behavior, focus of attention, thinking, imagery, image, feeling, sensation) are elicited in this situation? Which of them came first? How did this first activity/experience influence the other activities/experiences?”

Differences between MPC and essentialist explanations

Several features of the MPC type CT explanation should be noted and contrasted to those of the essentialist SET explanation. First, the CT model (causal schema) provides an explanatory network of causal relations within each symptom. That is, the CT model explains particular SAD symptoms by suggesting micronetworks (mechanisms) at the level of momentary states and behaviors. The causal connections between microlevel observables are in turn explained by reference to basic psychological principles—for instance, principles of limited attentional resources and of incompatibility of perceptual projects. Thus, whereas the CT model moves from symptoms to a more concrete level, the SET model moves in the opposite direction to a higher abstraction level, explaining symptoms by a latent construct. Second, the CT model explains the co-occurrence of symptoms by the dynamics of the micronetworks of single symptoms. Moreover, co-morbidity is explained by “bridge” symptoms—that is, symptoms that are common across disorders (Borsboom & Cramer, 2013). For SAD and panic disorder, for instance, panic attacks are a common symptom. Symptom activation may spread from one disorder to another through such “bridge” symptoms. These conceptualizations differ from the ones of PDT, where a common causal internal conflict (and/or ego deficit) explains the co-occurrence of symptoms as well as of disorders (Leichsenring et al., 2013). Third, different processes may explain the different symptoms within the same disorder. For instance, the elicitation of and exposure to less friendly behaviors from others contributes to in-situation social anxiety but has no role in anticipatory anxiety. In SET, on the other hand, all symptoms within a disorder are explained by the same cause, namely the latent conflict. Fourth, the same processes or activities—for instance, worry, rumination, and threat monitoring—may contribute in different networks across disorders; thus, these processes are transdiagnostic (Harvey, Watkins, Mansell, & Shafran, 2004). In SET, a latent conflict is considered transdiagnostic (Leichsenring & Salzer, 2014). Fifth, CT consists of facilitating insight into the individual patient’s particular micronetwork and helping the patient explore and manipulate the central causes in this network. SET consists of facilitating insight into the core conflict (expressive interventions) and also substituting some unfortunate behaviors and internal activities with more healthy ones (supportive interventions).

Requirements for establishing mechanisms of change

In the present examination of the extent to which CBT and PDT models can be subjected to a study of mechanisms of change, we will follow the well-established requirements developed by Kazdin (2009), ³ plus one that we will introduce here. One requirement is demonstration of a strong association between the psychotherapeutic intervention and the hypothesized mechanism of change as well as between the proposed mechanism and therapeutic change. Another is experimental manipulation—that is, to manipulate the proposed mechanism and show the impact on outcome. A third is to establish a timeline—that is, to demonstrate a time ordering of the proposed mechanism and outcome. Kazdin (2009) also emphasizes plausibility or coherence—that is, the need for logical conceptual models to explain those processes that are responsible for therapeutic change. These models should not only specify critical concepts but also how these operate to produce symptom change.

A fifth requirement not discussed by Kazdin (2009) is that within-person in contrast to between-person relationships must be examined. Therapists are primarily interested in how changes in a supposed mechanism (e.g., catastrophic cognitions) may cause change in an outcome variable (e.g., panic attacks) within one and the same patient (Hoffart, 2016). Still, most mechanism research to date has focused on between-person relationships—that is, the extent to which patients who change less than average on a potential mediator variable (e.g., catastrophic cognitions) also change less than average on an outcome variable (e.g., panic attacks). So it is the individual’s relationship to other individuals that is the focus of the analysis. However, this relationship is of more secondary interest to therapists who wish to achieve improvement within a particular patient. Methodologists have recently pointed out that between-person and within-person relations are different conceptually and empirically and that confounding or misattributing them could lead to faulty conclusions (Borsboom et al., 2003; Curran & Bauer, 2011; Hoffman & Stawski, 2009; Molenaar, 2004). Molenaar (2004) theoretically and empirically showed that results from between-person analyses can be generalized to within-person findings only when certain strict assumptions are met. First, each subject in the population has to obey the same statistical model (homogeneity of the population). That is, the main features of a statistical model describing the data (e.g., number of factors, factor loadings) are stable across subjects. Second, the data must be stationary—that is, have stable statistical characteristics (e.g., factor loadings) across time. No wonder that it seems to be the rule rather than an exception that within- and between-person relations are different. That is, between-person findings cannot be used as evidence that working with a given patient to change the presumed mechanism will improve the outcome for that particular patient. This implies that the concepts and propositions of therapeutic theories must be able to represent such within-person relationships.

Definition of the mechanisms in the two therapy models

Before examining whether the two models can be subjected to the required tests (Kazdin, 2009), we need to define their proposed change mechanisms. According to the SET model, SAD is caused by the patient’s particular CCRT and change in SAD is achieved through a change in the CCRT. That is, the CCRT is the mechanism through which SET is supposed to work. The gains from successful SET should be that the features of the CCRT, strength of the wish for affirmation and certainty of the expectation of humiliation, and thus the conflict, should be weakened. In addition, the ability to balance and negotiate these features should improve as more relationship options are opened to thought and expression (Crits-Christoph & Luborsky, 1990, p. 136).

The mechanistic changes proposed in the CT model of SAD may occur at different stages of the dynamic networks. The patient may react initially to social situations as something positive and have no anxious thoughts and images. If these still occur, he/she may relate differently to them, question their validity, or react in a detached way. If the anxious thoughts are not corrected or made distant, he/she may refrain from engaging in self-focused attention and safety behaviors and engage more in the social interaction. In short, successful CT therapy means lower levels on the causal variables of mini-networks and higher levels on complementary alternative variables. Moreover, it means that weaker connections between the observables are established.

Demonstrating a strong association between the interventions and the proposed mechanism

An empirical association between an intervention and change in a mechanism presuppose that these two entities are distinct. When two variables are not distinct, their relationship is tautological and not empirical (Edwards & Bagozzi, 2000). According to the cogent distinctions drawn by Doss (2004), the interventions and the patient’s change processes occur within therapy (or as homework). The mechanisms, on the other hand, play out in daily life outside of the therapist’s control. It is an open empirical question whether a disposition developed within the therapeutic context will extend to daily life. For instance, whether the occurrence of a weaker and more flexible CCRT pattern in relation to the SE therapist will lead to a weaker and more flexible CCRT in outside relationships is a purely empirical question. And less use of safety behavior in social tasks planned within CT may or may not spread to daily social situations. Thus, the requirement of independence between interventions and mechanism is fulfilled by both therapeutic theories. Examples of studies addressing this requirement are Connolly Gibbons et al. (2009) and Hoffart, Borge, and Clark (2016). Connolly Gibbons et al. found that a mechanism representing PDT—self-understanding of interpersonal patterns—increased more in psychodynamic than in CTs in a diagnostically heterogeneous sample receiving various treatments (one of these was SET). Hoffart et al. found that a mechanism derived from the cognitive model of SAD—estimated cost of negative social events—decreased more in CT than in IPT for SAD.

Demonstrating a strong association between the proposed mechanism and therapeutic change (outcome)

To what extent are the concepts of a conflict between a wish for affirmation and a fear of humiliation (CCRT) and of SAD distinct and separable? The meanings of these terms are fixed by the rules for their use in everyday language (Maraun & Peters, 2005). For instance, the rules for the emotion term fear may be stated as follows: Some object, person, event, or state of affair is perceived by a person P as a threat and this perception conveys motivation for P to engage in escape or avoidance behavior (Bergner, 2003). In the case of Peter, other persons are perceived by him as a threat of being humiliated, and this provides reasons for avoiding attention from them and avoiding appearing inept in their presence. The described behaviors constitute the criteria for the concept of expectation of humiliation. Most importantly, the behavioral criteria for the presence of this expectation are identical to some central symptomatic behaviors in SAD.

Similarly, the concept of wish for affirmation means that other persons’ affection and favor toward a person P is perceived by P as not (sufficiently) present, but P would like it to be (more) present. This conveys motivation for P to engage in attention-seeking, disclosing, and/or showing off behaviors. These behaviors constitute the criteria for the concept of wish for affirmation. The criteria for the concept of conflict between this wish and the fearful expectation are some compromises between the behavioral criteria for the single states. For instance, if our invented patient Peter suffered from this particular conflict, he would show a definite pattern of anxious reactions in SET. When the therapist asks him to talk about a difficult REP, poses clarifying questions, and reflects the content of what he says, he could show his concurrent wish for affirmation and fear of humiliation by experiencing feelings of anxiety, showing stiffness in body posture, trying to guess what answers the therapist would like to hear and give these, exhibiting a persistent concern of not giving good enough answers, eagerly confirming the accuracy of the therapist’s reflective comments, and reacting with shame and looking down if not able to provide answers. The wish without the fear would have manifested in authentic disclosing behavior. The fear without the wish would have manifested in silence and withdrawal. Conflicting goals, on the other hand, are behaviorally combined and balanced. Anyhow, the mental concepts of wish, expectation, and their conflict are defined by certain behaviors in particular circumstances.

Sloman (2005, p. 22) delineates what distinguishes causal relations from definitions. A definition identifies a word or phrase with a set of conditions in natural reality. If the conditions are met, then the word or phrase applies (i.e., the conditions are sufficient). Likewise, if the word or phrase applies, then the conditions are met (the conditions are necessary). If an object is geometric and has three sides, then it is a triangle. And if it is a triangle, then it is geometric and has three sides. In our case, if a person’s mental state is an expectation of humiliation, other persons are perceived as a threat of humiliation, and this provides reasons for avoiding attention from them and avoiding appearing inept in their presence. In definitions, linguistic entities are associated with conditions in natural reality. Causal relations, on the other hand, associate events in natural reality with other events in natural reality.

From the above, it is clear that the CCRT cannot be separated from the SAD symptoms. The fearful expectation phrase, which refers to a part of the CCRT, is defined in terms of socially anxious behavior directed at not drawing attention from others and not appearing inept to them. Thus, there is a logically necessary relation between the mental concept and the behaviors, and a proposition of them being related is a tautology. Likewise, the terms referring to a conflict between a wish for affirmation and fear of humiliation are defined by the anxious compromise behaviors described above. Consequently, wishes, expectations, and their conflict cannot be empirically identified without identifying the defining symptomatic behaviors. The relation between the mental concepts and the symptomatic behaviors is linguistic and logical and not causal and mechanistic. Thus, the proposition that CCRT change mediates symptom change lacks the hallmark of empirical propositions that their denial is logically possible (Bradley & Schwarts, 1979). Consequently, it cannot be refuted by experience. SET’s attribution of a wish for affirmation to the socially anxious beyond the obvious wish of not appearing inept is to apply an additional concept to and thus provide a new and extended description of the person’s behaviors. It is not to provide an explanation of them in terms of causes.

The problem of separating mechanism and outcome should show up in empirical studies of PDT. Of studies reviewed by Leichsenring and Salzer (2014), we found only one that accorded with the distinctions between intervention, patient change processes, and change mechanisms applied here. Connolly Gibbons et al. (2009) found that improvements in self-understanding from intake to termination were predictive of change in depression during the same period. Self-understanding was defined to include ability to recognize patterns and consider alternatives when they are experienced. However, self-understanding in this sense and depression cannot always be independently identified. Luborsky (1984, p. 97) describes a depressed woman with the relationship pattern that she first tells with enthusiasm about her involvement in activities, then she feels that the listener is losing interest, becomes depressed, and slows down speech. Thus, increased self-understanding of this relationship pattern—involving ability to consider alternatives when this pattern is experienced—would be indicated by a reduction of the depressive symptoms. Thus, the correlation between increased self-understanding and a decrease of symptoms of depression is not between independent phenomena and must be considered pseudoempirical.

Some may here object that we have exposed a wrong understanding of wishes and emotions. Bergner (2003) points out that a common way to think about the concepts of wishes and emotions is that they designate certain sorts of inner experiences—something that may be observed solely by the person who has them. In this view, terms such as wish, fear, and anger stand for unique, discriminable, private experiences. Their publicly observable manifestations in bodily expressions and behaviors are only empirical correlates or effects of their inner “essences.” In this view, which fits well with an essentialist model, wishes and emotions are independent of their behavioral manifestations. Similar to assessing whether a piece of metal is real gold by checking its number of protons, one could think that a final validation of the CCRT formulation would be a detection of an inner force—an experienced wish for affirmation—in connection with social anxiety. However, this view that wishes and emotions terms stand for inner experiences represents a logical impossibility, most profoundly shown by Wittgenstein (1953). Use of language builds on rules for what “the same” is—for instance, on rules for what is the same emotion in different persons and in the same person at different occasions. Without linguistic rules, verbal human communication could never be established. For instance, if a person was described as having an emotion at one occasion but not on another, we could not know whether the emotion or the meaning of the emotion term had changed. To at all be able to learn these rules and correct each other in their use, they must involve public and sharable phenomena (Maraun & Peters, 2005). Thus, the objection that emotion terms stand for private experiences is not valid. ⁴

The causal schemas of CT, on the other hand, provide micronetworks of causal relations between circumstances, activities, and experiences. Each network explains a single symptom of a disorder. These are observable phenomena that—either by behavioral observation or introspection—can be followed with attention and identified independent of each other. For instance, to test the CT model of SAD, Davis et al. (2016) repeatedly measured attention bias by reaction times in an experimental paradigm during a CBT program and related these data to self-reported anxiety. Due to their independent assessment, their correlation is meaningful.

Establishing a time ordering of the proposed mechanism and outcome

As wishes and emotions cannot be identified independently of their behavioral criteria, it gives no meaning to ask whether the wishes and emotions of the CCRT precede the explained symptoms. The requirement of demonstrating a timeline between the CCRT and outcome cannot be investigated. By contrast, the observable phenomena of the CT network can be identified independent of each other and—because they have genuine duration—tagged in time. Therefore, a timeline between them can be established. For instance, Davis et al. (2016) secured time ordering of supposed mechanism and outcome by relating attention bias scores a week to anxiety scores the subsequent week.

Experimental manipulation of the proposed mechanism to show the impact on outcome

Again, as wishes and emotions are identified through the observable reactions expressing them, it is logically impossible to remove the wishes and emotions of the CCRT without removing the symptomatic reactions. The observable phenomena of the CT network, on the other hand, can be identified independent of each other and manipulated separately. Thus, a putative mechanistic observable can be affected and the potential impact on outcome observed. For example, Woody (1996) showed that experimentally intensifying self-focus during a speech task increased the anxiety level of patients with social phobia. In a study of nonclinical adults, use of safety behaviors and self-focus were manipulated and demonstrated to be unhelpful in a number of ways (McManus et al., 2009). More examples of experimental studies of the mechanistic observables in the CT model of SAD are reviewed in Clark (2001).

Plausibility

Plausibility is provided to the extent that a model specifies how critical concepts operate to produce symptom change (Kazdin, 2009). This means that the processes linking the cause and the symptoms are specified. However, the cause in reflective models is construed as an unobservable intangible entity. The endeavor of finding a concrete process in between a “no-thing” and something observable (the symptoms) is impossible (see also Schmittmann et al., 2013). Therefore, the essentialist SET explanation cannot contribute to plausibility. The CT model, on the other hand, explains how concepts operate to produce symptom change and thus provide plausibility in two ways. First, the maintenance and change of single symptoms are accounted for by showing the interplay of their constitutive inner and outer activities and experiences. For instance, self-focus maintains fearful thoughts and thus social anxiety, which in turn provides positive feedback to self-focus, implying that self-focus is a promising target of change. Second, the influence of one symptom on another is explained by activities and experiences that “carry over” from the occurrence of one to the occurrence of another symptom—for instance, how anticipatory anxiety creates negative perceptual biases, which makes the person more likely to perceive signs of his/her own failure in the anticipated situation. The fact that this effect is covered by established psychological principles (i.e., confirming bias; Nisbett & Ross, 1980) contributes to further plausibility.

Ability to represent within-person relationships

The SET model proposes that a latent entity—the core conflict—causes the observable symptoms. Borsboom et al. (2003) have problematized a causal interpretation of the reflective relation between a latent construct (e.g., core conflict, depression) and its observable indicators. They show that it can be formulated only on a between-persons level. On this level, the reflective relation can be stated as follows: Population differences in position on the construct cause population differences in the expected responses. Thus, if person A has a stronger conflict than person B, A is also expected to have stronger symptoms than B. This statement fits common accounts of causality—for instance, the following three criteria: X can be considered a cause of Y if (a) X and Y covary, (b) X precedes Y, and (c) with other things the same, Y does not occur if X does not occur (Borsboom et al., 2003). In our example, we have that the degree of conflict and the strength of the observable symptoms co-vary, the conflict is supposed to precede the symptoms, and if there is no difference in position on the conflict dimension, there is no difference in the strength of the symptoms.

The individual within-person account of causality, however, is problematic. Consider the general statement of within-person causality: Within person A, variation in a presumed cause is related to variation in his/her reactions (e.g., item responses in a test, clinical symptoms). For a reflective relation, this would be as follows: Person A’s position on the construct caused his/her reactions. However, this statement is meaningless because in a reflective model, A’s position on a construct is conceptualized as a constant (Borsboom et al., 2003). If multiple and fluctuating behaviors and states are to serve as indicators of a person’s position on a latent depression construct, one must assume that this position is constant through the targeted measurement period (e.g., the preceding week). However, a within-person constant cannot be a cause of within-person variation. Thus, although reflective constructs such as depression and core conflict can be said to cause observable symptoms in a between-persons sense, it is meaningless to assert that a particular patient’s core conflict or depression caused his/her symptoms. To illustrate this point, consider the assessment of depression by the BDI-II. Here subjects are asked to rate the intensity and/or prevalence of various symptoms during the preceding week. The sum score indicates the latent value of depression for that week. An alternative way to measure this value would be to administer the BDI at the end of each day (see Fig. 3). The best estimate of it would then be the sum of the averages of every BDI symptom score across the 7 days. For the individual person, this constant value cannot explain day-to-day variation in symptoms, say, in lack of sleep. As shown in Figure 3, latent depression causes and explains only the mean value of a symptom across the days and not the symptom’s variations around this mean. On the other hand, through this way of measuring, we could estimate the within-person relationships between the symptoms—for instance, how a person’s night-to-night variation in lack of sleep influenced fatigue in the same person on subsequent days (upper part of Fig. 3).

OPEN IN VIEWER

Fig. 3.

A latent variable model of depression. The horizontal axis on the inserted graphs represents days of the week, and the vertical axis represents the item score level on the BDI-II. Latent depression influences the mean levels of depressive symptoms during the week, symbolized by the straight arrows. However, latent depression is unrelated to the time-specific variations of symptoms around these mean levels. On the other hand, daily variations in one symptom (e.g., lack of sleep) may relate to subsequent variations in another symptom (e.g., fatigue). The curved arrows represent the sequential relationships between symptoms. For simplicity, the connection from only one weekday to another is drawn. Due to the repeated measurement of symptoms within persons, these sequential relationships can be observed at the level of the individual person.

Essentialist SET explanations specify relationships between an abstract intangible entity (core conflict) and symptoms and say nothing about the relation between observables. Therefore, from such explanations, it is not possible to derive propositions like the following: “Within person A, if the observable (activity/experience) a decreases, then the observable (activity/experience) b will decrease.” Thus, as SET explanations say nothing about the relation between observables, interventions in SET addressing observable activities and experiences relate to a theoretical vacuum and lead to a neglect of evidence-based relationships between observables. For instance, the previously described supportive intervention of SET—fostering an encouraging inner dialogue—addresses the level of activities and experiences. Although making intuitive sense, it is not based on a causal analysis. That is, it is not examined how the inner dialogue affects and is affected by other activities and experiences such as feeling of anxiety. Actually, knowing the well empirically established relationship between the use of safety behavior and social anxiety (McManus et al., 2009), such inner dialogues may actually serve as safety seeking and thus carry the risk to consolidate the anxiety. Thus, the lack of a within-person causal perspective may have detrimental effects. Other supportive interventions such as exposure to feared situations align with well-established principles and are likely to have positive effects on symptoms. In any event, the lack of a within-person causal perspective in the SET model leaves it dependent on principles from external sources as a basis for the supportive interventions.

Essentialist models conceive of latent variables as sources of between-person differences but abstract away the processes taking place at the level of the individual (Borsboom et al., 2003). Thus, latent constructs point to causal entities that differ between individuals but do not vary within individuals. Therefore, latent constructs are not able to represent causal relationships between observable processes going on at the level of the individual.

MPC explanations and network models, on the other hand, are precisely about how variation in observable variables (e.g., mood, focus of attention) within a person relates to variation in these variables within the same person at subsequent time points. We have already mentioned one study of the temporal dynamics of depression symptoms (Bringmann et al., 2015). Another study examined the temporal dynamics within the symptom of social anxiety in SAD, finding that weekly changes in self-focus, estimated probability and estimated cost of negative social events, and safety behaviors predicted subsequent social anxiety (Hoffart et al., 2016). Moreover, social anxiety also fed back on some of these variables. This study had not enough data points per individual to examine the within-person relationships of single individuals. This was the case, however, in a study of Snippe et al. (2015). Six women with depressive symptoms filled out diary questionnaires during a mindfulness-based treatment. Day-to-day changes in mindfulness (being aware of present-moment experiences without judging) or repetitive thinking preceded day-to-day changes in depressive symptoms in some of the participants. Thus, because they focus on time-varying observables, the MPC type CT explanations own the required ability to represent within-person relationships.